[GRASS-dev] Band references need a better name and definition

Maris Nartiss maris.gis at gmail.com
Tue Sep 14 22:24:54 PDT 2021


Cross-posting for those who do not get notifications from GH. The
discussion is happening in GH:
https://github.com/OSGeo/grass/issues/1868
As the issue needs to be solved before 8.0 release, it would be nice
to get the feedback till 26th of September.
------------------------------------------

The issue

The term band reference was introduced with #63. There, perhaps, the
word reference was referring to the fact that the text is referring to
the band description stored in the file g.bands gives access to.
However, now an arbitrary band reference can be stored for any raster
map, i.e., it is more general than just bands from a predefined (or
possibly in the future, registered) list of bands associated with
particular sensors.

The band references can now be used (optionally assuming #1866) to
label any rasters (raster maps or bands) in an imagery group so that
signatures are associated with given set of bands rather than just a
fixed set of raster data.

Going even further, the usage does not have to limited to image
processing. For example, in modeling, standardized names such as
topographic__elevation and sea_surface_air__temperature are used to
describe inputs and outputs of models (see Landlab Standard Name
Definitions and CSDMS Standard Names).

Overall, band references quickly outgrew their original definition and
there is a potential to have good name suggesting the potential use.
At the same time, there is a potential for ending up with a cryptic
term which is difficult to explain and difficult to link to its
application.

The issues with the current name are:

The word band, although often used quite generally, might be tight too
much to remote sensing and may overlap with other usages of the term.
The word reference is the cryptic part. Name, label, band and id are
words people may have different ideas about, but they generally be
somewhat right. Reference, on the other hand, is a very special term,
loaded for some, meaningless for others.
The term band reference seems to be too long, resulting in modules and
functions using different combinations of band and bandref in the
interface instead of full band reference. Notably, bandref is not a
word. The word band by itself may be too ambiguous. Interestingly, I
don't think there is reference used as the only word, although that's
really the noun in the term.

Since the current system is not part of any release yet, now is the
time to get it right.

Naming options

band label
band name
band tag
band reference
band id
band identifier
semantic label
semantic name
semantic tag
standard variable name
standard field name
...

Expected resolution

We decide on a name for what these things are and even if it stays as
band references we are sure that that's how we want to call them.
The usage is consistent in documentation and in the interface (option
names, function names).

Additional context

We discussed this during the last call, but decided to open this as an
issue to discuss this in written form and asynchronously.


More information about the grass-dev mailing list